Klipper Musings

[TobyCWood]

I went round and round. Watched vids, read docs, Read tonsoposts.

Klipper in a nutshell...  

It's a way to get around the logjam, bottleneck of the current controller board processor architectures so one can go as fast as possible. One description which IMO was the most apropos was... it's a way to get to what it will be like when controller boards finally get multi-core processors.

To get there you need to build a system which IMO breaks some of the fundamentals of good system design as in decoupling and cohesiveness.

Besides the RPi as the master you'll need to flash Klipper to your printer's controller board which essentially couples it or slaves it to Klipper. If you ever need to use the printer as it was before you flashed it to accept Klipper's commands you will need to reflash the controller to do so.

In my digging I have seen writings which make it clear to me that unless you are going for speeds well above 150mm/s then you are probably not going to get all that much of a benefit for all your work in getting a Klipper based system working. So Klipper is GREAT for Vorons, HevORTS, RatRigs and Railcores. All CoreXy screamers. Mendles? No. Deltas? Not sure about that but it should be yes.

[TobyCWood]

I should add...

While it does what it intends to do... that is... go fast... it does it by ignoring some System Eng fundamentals which in turn has some significant downsides. Such as, reliability, maintainability, durability, etc. As such I would NOT recommend using it for machines that are used for production at any level. Nor would I recommend it for a those who prints things for fun occasionally. The only user/use cases I'd see it fitting with are those who are working to get their machines to go as fast as physically possible. Thats great for that use case. 

[Alan B]

I respectfully disagree with this view on Klipper.

Some would say the Klipper approach is a better overall design for a computing and control architecture, separating the low level hard realtime controls from the high level planning. I've seen tests that show processor overloads that affect printing occur even in our Prusa's with Marlin at normal print rates with some design files. Most industrial CNCs use low level hardware or firmware and sub-processors and separate the controls into layers in similar fashion to Klipper's architecture. Overall performance often requires layering a system and splitting up the processing in an intelligent fashion. These days most systems are moving to parallel processing to improve performance and maintainability. In the case of Klipper one of the benefits of the architecture is the easy development of high level functionality that is taking place. Just throwing a bigger processor at a problem may seem easy but the reality of scheduling and synchronization to guarantee meeting the low level real time requirements within the same CPU becomes difficult and frequently unstable in the long term. Dedicating a separate processor to this simplifies the overall maintenance and development. The high level developers don't need to be concerned with the low level realtime hardware or software. The low level time critical processing requires the most skilled developers, and once that is done it generally doesn't have to be fiddled with. Having a reliable and stable low level separated hardware/software base is good for the project, and having it on a physically separate processor enforces useful software communications boundaries. When the low level code is running on a high level CPU core it is subject to many influences that can derail it and require more highly skilled development and debugging much later in the project. Each new OS or compiler release threatens it.

One existing successful example of production 3d printing with Klipper is the Voron Print It Forward program. This group of folks are running farms of mostly Klipper based Vorons producing parts for the new Voron builders worldwide. They are definitely in production mode. Some use 32 bit RepRap controllers but most use Klipper based systems, and if the Klipper systems were problematic compared to the RepRaps we'd see it in the discord right away.

I haven't run Klipper yet, I expect there will be some frustration getting a system going as with anything new. Reflashing back to Marlin is pretty trivial if one decides to. Probably a lot easier than recabling a new 32 bit controller that likely has different connectors.

My biggest concern about running Klipper on a Prusa is potentially stepping away from a lot of useful manufacturer and community support. So instead I'll use Klipper on a Voron to gain experience with it - where the community support is strong. The Railcore is predominantly 32 bit RepRap community, so again not perhaps the best place to learn and use Klipper.

The LDO Voron kits are bringing Prusa level quality to the Vorons in kit form. That is going to explode when LDO comes to market with the Trident and 2.4 models in the first half of next year. We'll see a major uptick in Klipper as those products come to market.

It remains to be seen what Prusa will do with 32 bits on the MK3. They did put it on the Mini, and clearly they are headed that way with the XL (and print farm) and probably the Mk4 whatever that turns out to be. But Prusa seems to be lagging compared to LDO.

Interesting times...

--
You received this message because you are subscribed to the Google Groups "3D Printing Tips and Tricks" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3d-printing-tips--...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/76182cff-33f6-4a81-b730-5e0f951723e5n%40googlegroups.com.

[David Mason]

I came here to right most of what Alan had said,but you put it well so no need to repeat it. Klipper is in my mind a better solution overall. But Andy you are right in where I'd use it personally with one exception that I'd probably standardise an ender 3 just to match my work flow for my better printers if I owned one. 

I would also agree that where there is very good customer support from the company its worth while sticking to their system. Prusa are the only lot I can think of that probably meet that standard though. 

One key bit that I feel like neither of you covered is the modular expandability of klipper. Want to convert your existing printer to a tool changer or mmu? Buy another low cost control board and use all 5 stepper drivers from that for the extra axis. Want a whole load of led control? Buy the klipper led expansion board that the community created and China make. But again this puts the use case in higher end or heavily modified printers. 

[LukeH]

Klipper was/is designed as a fix to issues that were holding back 3D printer design maybe four or five yers ago, where most printers were powered by crappy 8-bit control boards (like the Prusa printers still are), using outdated Mendel designs (like the Prusa still are), and printing at ridiculously slow speeds (like the Prusa still are).

Klipper was a revelation - using the Raspberry Pi that people were adopting to run OctoPrint to also do the bulk of the computing grunt work meant that the system could calculate more steps per second, and print much faster (well, have the potential to print faster depending on your printer design). Then 32-bit processors came to printer control boards in a big way (unless you are buying a Prusa, I doubt you can even buy a printer with an 8-bit board these days), which are mostly plenty fast enough for the task for basically everyone except those who are attempting to print a terrible quality Benchy in less than six minutes. 

Since that time Klipper has added features that show real promise, such as the ability to use multiple control boards (although the number of 3D printers that need more than the 5 or 6 stepper drivers on most boards these days is pretty small, and you can now buy expansion boards for many decent control boards, or even just board with 8 or 10 stepper drivers). Control boards these days also offer a full integrated package (WiFi, inputs for USB mass storage, web interfaces, etc), which eliminates the need to run another single board computer (like a RPi), meaning that although Klipper is available at no cost, people wanting to use it may not have a RPi already connected to their printer, and so there will be a hardware expense.

Input shaping though is impressive, and probably worth the price of admission.

Klipper is almost certainly behind the rise of the CoreXY design, and the success of designs like Voron, RailCore, HEVO, etc.

Klipper’s biggest drawback though it that it remains targeted at enthusiasts, and so hasn’t achieved a more “mainstream” (as if anything in 3D printing is mainstream) appeal. Too much focus on features, and not enough on interface and ease of use IMO.

The sad part is that slicers haven’t kept up - while we have these 32- and 64-bit computers running our printers, slicers still don’t natively implement things like GCode commands (G02 or G03) for arcs or other features that would require a higher level of computing grunt from the control board than what is available on 8-bit boards, even though those are now few and far between. There are gains in print quality that slicers could unlock beyond what Klipper could provide, but I’m guessing it would take a significant refactoring of code, which is difficult for open source projects.

[bkaufm...@gmail.com]

I went from Marlin on all my machines to Klipper, first on a Railcore, now on everything, including an Ender 3. It has enabled me to get to 80 mm/s comfortably on what should be a 50 mm/s printer. Obviously, it enables a lot faster printing than that on a railcore. There are people that have used Duet boards with Klipper. The biggest advantage there is just input shaping, but that is apparently in development for RRF, so it may not be an advantage for long. It isn't just the 32 bit, it is the motion planner that is just plain better than average. For me, it was that Klipper yielded better prints than I could get on Marlin. Might not be the case now, but it was at the time. RRF has a good motion planner, so less of an advantage. 

I don't know that Klipper has a drawback, it just has characteristics of its users. I wouldn't feel uncomfortable running klipper on a production-oriented prosumer printer. Wouldn't feel uncomfortable with RRF either. Marlin - IMHO that is a mistake. The challenges with flashing and maintenance are a deal breaker. 

[Dushyant Ahuja]

I'd like to add one more aspect - it's flexibility. I'm using Klipper on my CoreXY, as well as my custom 2D Plotter. For my 2D plotter, I can easily reprogram what specific gcode commands do, so that I can easily switch my processes. E.g. I can use SuperSlicer to slice an STL and use the retract/unretract commands for pen up and pen down, or use lightburn to convert an SVG, and use laser on/off for pen up/down. Plan to use it on my sand table soon - not that it needs Klipper. 

And it's a lot more fun to use. 

To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/ad8b1844-753b-4da9-9da9-bc0b4b1741c0n%40googlegroups.com.

[TobyCWood]

Ok... I'll add one more aspect myself.

So ya like to go fast? Thats cool. You can get your HW and your SW to make it fast and you can get Input shaping to cancel out much of the movement in the gantry... but... what you can't do is control the momentum of the melted polymer that you are extruding. The stuff is liquid and moving. Think about it.

Can you control it? Yeah... you can print thinner layers which in turn increases print time significantly as well as adds tighter tolerance requirements.
Is this only possible with Klipper? Not at all! I've been taking the speed to the limits of the polymer momentum with a Duet and RRF. A cohesive design which completely decouples all the settings as gcode.

[david mason]

I'm a huge duet fan too. My vcore3 runs duet at the moment. The custom build that I took apart to turn into the vc3 ran that same duet board for the last year or more.

I certainly aren't saying there are not other great options available. You just seemed to dismiss klipper as a poor solution too easily. People have made all the points needed as counter arguments. I'd like to see you give it a prolonged test run before sayings it's a waste of time(I'm not aware that you have, based on your podcasts, please correct if wrong). 

Get Outlook for Android

---

From: 3d-printing-...@googlegroups.com <3d-printing-...@googlegroups.com> on behalf of TobyCWood <andyc...@gmail.com>
Sent: Saturday, November 6, 2021 4:50:38 PM
To: 3D Printing Tips and Tricks <3d-printing-...@googlegroups.com>
Subject: Re: Klipper Musings

 

You received this message because you are subscribed to a topic in the Google Groups "3D Printing Tips and Tricks" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/3d-printing-tips--tricks/Qu8LxDET1G0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to 3d-printing-tips--...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/60bb2c35-95df-4c75-a9bd-920da0c8c8b1n%40googlegroups.com.

[Luke Hartfiel]

Keep in mind though, that the ultimate theoretical goal of 3D printing would be layers so small you can’t see or feel them, printed so fast it only takes seconds to print large objects, rather than hours and days.

Klipper doesn’t get us there. None of the current firmwares do (yes - Klipper is firmware, or at least includes a firmware portion installed on the control board). The physics of liquid flows, heat transfer, and momentum might mean that we will never get there with current approaches.

Klipper does however help move us forward, which is good for everyone, even if you don’t use Klipper yourself. It helped moved us from thinking that 50mm/s was a good “normal” print speed, to now considering 150mm/s or 200mm/s to be a good “normal” speed in just a couple of years (and 50mm/s to be considered “painfully slow”).

On 7 Nov 2021, at 3:50 am, TobyCWood <andyc...@gmail.com> wrote:

Ok... I'll add one more aspect myself.

You received this message because you are subscribed to a topic in the Google Groups "3D Printing Tips and Tricks" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/3d-printing-tips--tricks/Qu8LxDET1G0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to 3d-printing-tips--...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/60bb2c35-95df-4c75-a9bd-920da0c8c8b1n%40googlegroups.com.

[TobyCWood]

Here’s my problem, imo it’s a waste of time to do a klipper setup for a Mendel. It’s got to be a corexy machine like a Voron or hevort As such it means building one and I got no room for another machine! But when forced…
What’s the best fully sourced Voron kit? Hevort and ratrig will cost too much.

On Thursday, November 4, 2021 at 10:26:57 AM UTC-7 TobyCWood wrote:

[Kurt at VRFX]

I will admit - I am joining this thread rather Late in the game (albeit
like 3 days late) - and there has already been a bunch of discourse.
But, I just wanted to hit on the one point - upon which I am replying below.

So, as already mentioned - Klipper is about speeding things up. But, in
regards to Deltas - they are Already way fast, in general, as opposed to
Cartesian based systems. So, in that regards - does it make sense to
apply Klipper to a Delta? Sure - it can MAYBE be made faster. But, is it
possible, that since a Delta is generally already a good bit faster -
that something like Klipper making it faster may also make problems that
happen with fast printing(like ringing and backlash) ever worse?

Well - maybe others will already have answered my question with their
replies. I will continue reading the thread now. Just got done reading
the VERY Interesting reply from Alan B.

L8r...
-Kurt

[Luke Hartfiel]

Kurt,

One of the impressive features of Klipper is Input Shaping, where you can eliminate ringing by calibrating the printer nozzle to oscillate at a frequency that cancels out the ringing (it actively changes the frequency for different print speeds, and print directions). There is an open-loop version, where you print calibration prints, measure the ringing, and set the settings, and a closed-loop version, where you install accelerometers on the printer, and it measures and compensates as it goes. It even apparently works on a Delta, although I think I’ll just wait for the closed loop version to come out of beta on RRF (due soon, I believe).

That means you can use Klipper to theoretically have faster prints that are actually better quality than a slower print on a different firmware without input shaping. Sort of a win-win situation, where everything is improved.

I would go as far as to say that input shaping is probably the single biggest innovation in 3D printing in years (at least since the Bondtech BMG).

> --
> You received this message because you are subscribed to a topic in the Google Groups "3D Printing Tips and Tricks" group.
> To unsubscribe from this topic, visit https://groups.google.com/d/topic/3d-printing-tips--tricks/Qu8LxDET1G0/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to 3d-printing-tips--...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/f6964e10-e8cb-ef68-0ad4-181abfa340cf%40optonline.net.

[Kurt at VRFX]

Interesting that you would mention those commands - as I posted about
the Arc Welder project on here a little while ago. We were trying to
experiment with the Arc Welder project and running those type of Gcode
commands on our Essentium printers - but, we were running into problems
with the Arc Welder project implementation...

L8r...
-Kurt

[Kurt at VRFX]

Luke - thanks for the feedback. I did indeed hear Andy talk about the
Input Shaping capabilities of Klipper in his Podcast - and, I have to
say - I was VERY Excited about it. So, yeah - I guess that could even in
theory help a Delta to go even FASTER and reduce the defects caused by
going So Fast!

L8r...
-Kurt

[TobyCWood]

As others on this thread have already noted, there is an assumed value to Klipper that goes beyond Input Shaping, that is it's distributed processing approach which allows for much more flexible solutions as well as it's decoupled way of doing settings. 

[david mason]

Cheapest fully sourced voron kits are the v0 ones. If space is an issue that also works in your favour as they are tiny!

Quality wise the LDO kit is the way to go - building mine right now and can't complain about a single part of the supplied kit so far. Although obviously time will tell. LDO kit is £600 gbp. The aliexpress kits go for around £400 I think. There is a discord where people are doing kit feedback which I can share if you'd like to brave the world of discord 😂

Get Outlook for Android

---

From: 3d-printing-...@googlegroups.com <3d-printing-...@googlegroups.com> on behalf of TobyCWood <andyc...@gmail.com>

Sent: Monday, November 8, 2021 5:17:23 AM

To: 3D Printing Tips and Tricks <3d-printing-...@googlegroups.com>
Subject: Re: Klipper Musings

To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/ec21b4a7-3a4a-4f1e-a931-63cfce838a30n%40googlegroups.com.

[TobyCWood]

OK... Im looking deeper and deeper... Thinking about the printer's controller board which has the MCU(now I get it). As such the selection of the controller board may be critical especially since so many of the posts on the Klipper Group on FB are about MCU command problems. Therefore it seems to make sense to pick a board with an MCU that is known to work with the Klipper MCU commands.
I dig around and find this table:

https://user-images.githubusercontent.com/8399317/56072885-d059b500-5d69-11e9-9ee4-43a69abd9d8a.png

uh-oh... 

Per this table it seems the only controller boards that get a "Good" rating are the 8 bit, older boards. Every MCU on this table which is 32 bit is rated as "Beta" ( at best) which in this case means..."All features implemented mostly working". DOH!

I can't see building anything and then putting in an older 8 bit board and I'm not all that into the idea of using an MCU that just "mostly" works!

[david mason]

The board the voron design team recommend in the BOM is a 32 bit board and there are a good few hundred v0 or v0 1 users in the wild using them mostly without issues. I suspect the documentation that you linked is out of date. An issue with all open source work. Everyone likes adding features but hates documenting it hah.

Bigtreetech SKR mini E3 v2 is the board FYI. I'm probably like 5 or 6 hours off being ready to test that specific board in terms of my v0 build. Not sure if allowed as it's a 3d sprinting blog (mine) but here is a link to my build series. I can . https://3dpandme.wordpress.com/voron-v0-1-ldo-kit-build-log/

The spider and octopus boards are widely used in the vcore3 and the voron builds. Both have lots of users and examples of working configurations out the box. Although there have been some hardware related issues with the early releases of those boards. 

Also good news some of these boards also run RRF (mellow super infinity for example 

Get Outlook for Android

---

From: 3d-printing-...@googlegroups.com <3d-printing-...@googlegroups.com> on behalf of TobyCWood <andyc...@gmail.com>

Sent: Monday, November 8, 2021 4:55:50 PM

To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/e6e642dc-ce02-4442-b9cb-07d867b6caa5n%40googlegroups.com.

[Kurt at VRFX]

Andy - I hear you. The feeling I got from some of the postings and discussions seemed to be that Klipper MAY have originally been developed specifically for 8-bit boards, to make them (or rather - 3DP's with them) run faster. Since, as we know - generally machines w/32-bit boards obviously will generally run faster than 8-bit board 3DP's. So, I think Klipper was developed with the intention of giving folks faster speeds on their 3DP's if they already rung a 3DP with an 8-bit board. And, yeah - the Input Shaping is definitely a Hot Topic for Klipper - as others at my co. have also been discussing Input Shaping.

Fun times in the world of 3D Printing indeed!!!

L8r...
-Kurt

You received this message because you are subscribed to the Google Groups "3D Printing Tips and Tricks" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 3d-printing-tips--...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/e6e642dc-ce02-4442-b9cb-07d867b6caa5n%40googlegroups.com.